pith. sign in

arxiv: 2606.29855 · v1 · pith:KXCJMY3Knew · submitted 2026-06-29 · 💻 cs.CV

RainODE: Continuous-Time Precipitation Forecasting with Latent Neural ODEs

Yeeun Seong , Doyi Kim , Minseok Seo , Changick Kim This is my paper

Pith reviewed 2026-06-30 06:37 UTC · model grok-4.3

classification 💻 cs.CV
keywords precipitation forecastingneural ODEcontinuous-time modelinglatent dynamicsstochastic refinementBrownian Bridgeradar dataadvective motion
0
0 comments X

The pith

RainODE models precipitation evolution in latent space with a Neural ODE for advective motion plus Brownian Bridge refinement for residuals, enabling sharp forecasts at any time interval.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that precipitation forecasting benefits from reformulation as a continuous-time dynamical system rather than discrete steps, using a latent Neural ODE to capture dominant large-scale advective motion while adding a stochastic module for intensity changes. A sympathetic reader would care because observational limits and computational costs make high temporal resolution difficult with standard discrete models. The combination allows derivative-consistent dynamics and arbitrary-time inference without over-smoothing. Experiments on SEVIR and the RAPID dataset show gains across intervals and regimes.

Core claim

Precipitation forecasting is reformulated as a continuous-time dynamical system and modeled in latent space using a Neural ODE that captures the dominant large-scale advective motion of precipitation systems. A purely deterministic ODE struggles with non-advective intensity changes such as localized growth, decay, and sub-grid variability, leading to over-smoothed predictions, so a stochastic source modeling module based on a Brownian Bridge formulation is introduced to refine residual intensity variations while preserving advective consistency. By combining deterministic continuous dynamics with stochastic refinement, the framework enables arbitrary-time inference while maintaining sharp pr

What carries the argument

Latent Neural ODE for deterministic advective dynamics combined with Brownian Bridge stochastic source modeling module for residual intensity variations

If this is right

  • Forecasts become possible at arbitrary time intervals without requiring dense discrete modeling or post-hoc interpolation.
  • Advective consistency is preserved while fine-grained structures are restored through stochastic refinement.
  • Performance improves consistently across multiple temporal intervals and precipitation regimes on SEVIR and RAPID.
  • The approach addresses both accuracy and temporal flexibility constraints in radar-based precipitation nowcasting.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The latent continuous formulation could support fusion with irregularly sampled satellite observations at varying intervals.
  • Similar ODE-plus-stochastic refinement structures might apply to other advection-dominated fields like cloud motion or pollutant transport.
  • Operating in latent space could lower the cost of generating high-resolution forecasts compared to full-grid discrete simulations.

Load-bearing premise

Dominant precipitation evolution is advective and can be captured by a deterministic latent ODE, with residuals adequately modeled by a Brownian Bridge that does not break the advective consistency.

What would settle it

If removing the Brownian Bridge module produces no measurable loss in sharpness or accuracy at intermediate times between radar observations on the RAPID dataset, while the full model fails to outperform standard interpolation baselines.

Figures

Figures reproduced from arXiv: 2606.29855 by Changick Kim, Doyi Kim, Minseok Seo, Yeeun Seong.

Figure 1
Figure 1. Figure 1: Task Description. Increasing temporal resolution causes (a) error accumu￾lation, (b) computational bottlenecks, and (c) temporally uncorrelated predictions. In contrast, (d) RainODE models shared continuous-time latent dynamics via a Neural ODE, enabling stable long-horizon forecasting and arbitrary-time inference. and consistency across outputs is not explicitly modeled. Nevertheless, most ex￾isting appro… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of RainODE. Input radar sequences are encoded into latent fea￾tures, which evolve continuously via a Neural ODE conditioned on the target fore￾casting interval. Consistency loss aligns ODE forecasts YˆODE with teacher predictions Yˆteacher, while reconstruction loss supervises Yˆteacher. The stochastic source modeling (SSM) module further refines predictions using a Brownian Bridge diffusion proce… view at source ↗
Figure 3
Figure 3. Figure 3: Temporal dynamics across RAPID. As the time interval increases (10-min, 30-min, and 60-min), the non-advective fraction ρ shifts upward, structural evolution plays a progressively larger role relative to displacement at longer horizons. The broader spread at 30 minutes suggests a transition between motion-dominant and evolution￾driven regimes. Temporal Dynamics Across Intervals. RAPID provides forecasting … view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative results on SEVIR and RAPID-60min. (top) SEVIR forecasts at +20, +40, and +60 minutes. (bottom) RAPID-60min forecasts at +2, +4, and +6 hours. Red dashed lines are overlaid to visually assess spatial alignment of predicted precipitation structures. Pixel values are encoded in [0,255], with higher values corre￾sponding to extreme rainfall [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Continuous-time forecasting up to 6 hours. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The robustness of continuous modeling. Performance is measured using CSI-16/160/219 at 10-minute evaluation intervals. RainODE is trained at 1-hour in￾tervals and evaluated at finer temporal resolution without retraining, while baseline models rely on autoregressive (AR) rollout for long-horizon prediction. The comparison highlights a key difference between discrete and continuous￾time forecasting. While S… view at source ↗
Figure 7
Figure 7. Figure 7: Temporal consistency of latent trajectories on RAPID-60min. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
read the original abstract

In precipitation forecasting, not only accuracy but also temporal resolution is critical. However, increasing temporal resolution is constrained by observational limitations and the computational cost of dense discrete modeling. To overcome this limitation, we reformulate precipitation forecasting as a continuous-time dynamical system and propose RainODE, a framework that models precipitation evolution in latent space using a Neural ODE. This formulation enables derivative-consistent temporal dynamics and captures the dominant large-scale advective motion of precipitation systems. Nevertheless, a purely deterministic ODE struggles to represent non-advective intensity changes such as localized growth, decay, and sub-grid variability, often leading to over-smoothed predictions. To address this issue, we introduce a stochastic source modeling module based on a Brownian Bridge formulation, which refines residual intensity variations and restores fine-grained structures while preserving advective consistency. By combining deterministic continuous dynamics with stochastic refinement, RainODE enables arbitrary-time inference while maintaining sharp predictions. Experiments on SEVIR and the newly introduced Radar-based Precipitation Integrated Dataset (RAPID) demonstrate consistent improvements across multiple temporal intervals and precipitation regimes. The code is available at https://github.com/SeongYE/RainODE.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The paper claims to reformulate precipitation forecasting as a continuous-time dynamical system via a latent Neural ODE that captures dominant advective motion, augmented by a Brownian Bridge stochastic source module to model non-advective residuals (growth/decay, sub-grid variability) without breaking consistency; this enables arbitrary-time inference with sharp predictions and yields consistent gains over discrete baselines on SEVIR and the new RAPID dataset.

Significance. If the deterministic-stochastic separation holds and the Brownian Bridge integrates without perturbing advective trajectories or derivative consistency, the framework would meaningfully advance continuous-time nowcasting by relaxing the need for dense discrete sampling while preserving fine structures. The public code release and introduction of the RAPID dataset are concrete strengths that support reproducibility and further work in the area.

major comments (1)
  1. [Abstract] Abstract: the central claim that the stochastic module 'restores fine-grained structures while preserving advective consistency' is load-bearing for arbitrary-time inference, yet the description provides no derivation or integration scheme showing how the Brownian Bridge is added to the Neural ODE without coupling back into the latent trajectory or violating the advective assumption; this separation must be demonstrated explicitly (e.g., via ablation on derivative consistency or trajectory smoothness) for the claimed benefits to follow.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and for identifying the need to make the separation between the deterministic Neural ODE and stochastic module fully explicit. We address the single major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the stochastic module 'restores fine-grained structures while preserving advective consistency' is load-bearing for arbitrary-time inference, yet the description provides no derivation or integration scheme showing how the Brownian Bridge is added to the Neural ODE without coupling back into the latent trajectory or violating the advective assumption; this separation must be demonstrated explicitly (e.g., via ablation on derivative consistency or trajectory smoothness) for the claimed benefits to follow.

    Authors: We agree that the abstract does not contain an explicit derivation or integration scheme, and that an ablation on derivative consistency and trajectory smoothness would strengthen the central claim. In the manuscript the Brownian Bridge is introduced as an additive stochastic source applied after latent ODE integration and decoding (Section 3), so that it does not enter the latent dynamics or the ODE solver. Nevertheless, the current text does not provide the requested quantitative verification of non-coupling. We will therefore (i) revise the abstract to reference the post-integration placement, (ii) add a short derivation of the integration scheme in Section 3, and (iii) include an ablation measuring the norm of dz/dt differences and trajectory smoothness with/without the bridge. These changes will appear in the revised version. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces RainODE as a modeling framework that uses a Neural ODE to capture deterministic latent dynamics for advective precipitation motion and augments it with a Brownian Bridge module for residual non-advective variations. This separation is asserted as a design choice to enable continuous-time inference, without any equations or claims in the provided text reducing the central result to a self-referential fit, parameter renaming, or load-bearing self-citation. The derivation relies on standard Neural ODE and stochastic process components with external dataset validation, remaining self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The model rests on the premise that precipitation dynamics separate into advective motion (handled by ODE) and residual intensity (handled by Brownian Bridge); no explicit free parameters or invented entities are named in the abstract.

axioms (2)
  • domain assumption Precipitation evolution is dominated by large-scale advective motion that can be represented in latent space by a Neural ODE.
    Invoked to justify the deterministic ODE component.
  • domain assumption Non-advective intensity changes can be modeled as residual stochastic processes without violating advective consistency.
    Basis for adding the Brownian Bridge module.

pith-pipeline@v0.9.1-grok · 5733 in / 1293 out tokens · 16634 ms · 2026-06-30T06:37:54.170815+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

46 extracted references · 6 canonical work pages

  1. [1]

    Expert Systems with Applications268, 126301 (2025)

    An, S., Oh, T.J., Sohn, E., Kim, D.: Deep learning for precipitation nowcasting: A survey from the perspective of time series forecasting. Expert Systems with Applications268, 126301 (2025)

    2025

  2. [2]

    Deep learning for day forecasts from sparse observa- tions.arXiv preprint arXiv:2306.06079, 2023

    Andrychowicz, M., Espeholt, L., Li, D., Merchant, S., Merose, A., Zyda, F., Agrawal, S., Kalchbrenner, N.: Deep learning for day forecasts from sparse ob- servations. arXiv preprint arXiv:2306.06079 (2023)

    work page arXiv 2023

  3. [3]

    IEEE Geoscience and Remote Sensing Letters19, 1–5 (2022)

    Bai, C., Sun, F., Zhang, J., Song, Y., Chen, S.: Rainformer: Features extraction balanced network for radar-based precipitation nowcasting. IEEE Geoscience and Remote Sensing Letters19, 1–5 (2022)

    2022

  4. [4]

    In: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)

    Chen, L., Du, F., Hu, Y., Wang, Z., Wang, F.: Swinrdm: integrate swinrnn with diffusion model towards high-resolution and high-quality weather forecasting. In: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI). vol. 37, pp. 322–330 (2023)

    2023

  5. [5]

    Advances in Neural Information Processing Systems (NeurIPS) 31(2018)

    Chen, R.T., Rubanova, Y., Bettencourt, J., Duvenaud, D.K.: Neural ordinary dif- ferential equations. Advances in Neural Information Processing Systems (NeurIPS) 31(2018)

    2018

  6. [6]

    International Conference on Learning Representations (ICLR) 2023 Workshop on Tackling Climate Change with Machine Learning (2023)

    Demiray, B.Z., Sit, M., Demir, I.: Efficienttempnet: Temporal super-resolution of radar rainfall. International Conference on Learning Representations (ICLR) 2023 Workshop on Tackling Climate Change with Machine Learning (2023)

    2023

  7. [7]

    Nature communications13(1), 5145 (2022)

    Espeholt,L.,Agrawal,S.,Sønderby,C.,Kumar,M.,Heek,J.,Bromberg,C.,Gazen, C., Carver, R., Andrychowicz, M., Hickey, J., et al.: Deep learning for twelve hour precipitation forecasts. Nature communications13(1), 5145 (2022)

    2022

  8. [8]

    In: Proceedings of the International Con- ference on Machine Learning (ICML) (2025)

    Feng, W., Li, X., Wu, Z., Lin, K., Yu, D., Ye, Y., Wang, Y.: Perceptually con- strained precipitation nowcasting model. In: Proceedings of the International Con- ference on Machine Learning (ICML) (2025)

    2025

  9. [9]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Gao, Z., Tan, C., Wu, L., Li, S.Z.: Simvp: Simpler yet better video prediction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 3170–3180 (2022)

    2022

  10. [10]

    Advances in Neural Information Processing Systems (NeurIPS)36, 78621–78656 (2023)

    Gao, Z., Shi, X., Han, B., Wang, H., Jin, X., Maddix, D., Zhu, Y., Li, M., Wang, Y.B.: Prediff: Precipitation nowcasting with latent diffusion models. Advances in Neural Information Processing Systems (NeurIPS)36, 78621–78656 (2023)

    2023

  11. [11]

    Gao, Z., Shi, X., Wang, H., Zhu, Y., Wang, Y.B., Li, M., Yeung, D.Y.: Earthformer: Exploringspace-timetransformersforearthsystemforecasting.AdvancesinNeural Information Processing Systems (NeurIPS)35, 25390–25403 (2022)

    2022

  12. [12]

    In: Proceedings of the International Conference on Machine Learning (ICML) (2024)

    Gong, J., Bai, L., Ye, P., Xu, W., Liu, N., Dai, J., Yang, X., Ouyang, W.: Cas- cast: Skillful high-resolution precipitation nowcasting via cascaded modelling. In: Proceedings of the International Conference on Machine Learning (ICML) (2024)

    2024

  13. [13]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Guen, V.L., Thome, N.: Disentangling physical dynamics from unknown factors for unsupervised video prediction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 11474–11484 (2020)

    2020

  14. [14]

    arXiv preprint arXiv:2406.04678 (2024)

    Kim, D., Seo, M., Choi, Y.: Ace metric: Advection and convection evaluation for accurate weather forecasting. arXiv preprint arXiv:2406.04678 (2024)

    work page arXiv 2024

  15. [15]

    arXiv preprint arXiv:2409.20117 (2024)

    Kim, D., Seo, M., Lee, H., Seo, J.: Masked autoregressive model for weather fore- casting. arXiv preprint arXiv:2409.20117 (2024)

    work page arXiv 2024

  16. [16]

    Technical Report (2020),https://meteofrance.github.io/meteonet/ RainODE 17

    Larvor, G., Berthomier, L., Chabot, V., Le Pape, B., Pradel, B., Perez, L.: Me- teonet, an open reference weather dataset by meteo france. Technical Report (2020),https://meteofrance.github.io/meteonet/ RainODE 17

    2020

  17. [17]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Li, B., Xue, K., Liu, B., Lai, Y.K.: Bbdm: Image-to-image translation with brow- nian bridge diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 1952–1961 (2023)

    1952

  18. [18]

    In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Lin, K., Zhang, B., Yu, D., Feng, W., Chen, S., Gao, F., Li, X., Ye, Y.: Alphapre: Amplitude-phase disentanglement model for precipitation nowcasting. In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 17841–17850 (2025)

    2025

  19. [19]

    In: Proceedings of the International Conference on Machine Learning (ICML) (2023)

    Liu, G.H., Vahdat, A., Huang, D.A., Theodorou, E.A., Nie, W., Anandkumar, A.: I2sb: Image-to-image schrodinger bridge. In: Proceedings of the International Conference on Machine Learning (ICML) (2023)

    2023

  20. [20]

    arXiv preprint arXiv:2410.06560 (2024)

    Liu, P., Zhou, T., Sun, L., Jin, R.: Mitigating time discretization challenges with weatherode: A sandwich physics-driven neural ode for weather forecasting. arXiv preprint arXiv:2410.06560 (2024)

    work page arXiv 2024

  21. [21]

    In: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)

    Ning,S.,Lan,M.,Li,Y.,Chen,C.,Chen,Q.,Chen,X.,Han,X.,Cui,S.:Mimoisall you need: A strong multi-in-multi-out baseline for video prediction. In: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI). vol. 37, pp. 1975–1983 (2023)

    1975

  22. [22]

    Weather and Forecasting31(1), 329–340 (2016)

    Otsuka, S., Tuerhong, G., Kikuchi, R., Kitano, Y., Taniguchi, Y., Ruiz, J.J., Satoh, S., Ushio, T., Miyoshi, T.: Precipitation nowcasting with three-dimensional space– time extrapolation of dense and frequent phased-array weather radar observations. Weather and Forecasting31(1), 329–340 (2016)

    2016

  23. [23]

    In: International Conference on Learning Representations (ICLR) (2026)

    Ribeiro, B.P., Pucer, J.F.: Flowcast: Advancing precipitation nowcasting with con- ditional flow matching. In: International Conference on Learning Representations (ICLR) (2026)

    2026

  24. [24]

    Mathema- tische Annalen46(2), 167–178 (1895)

    Runge, C.: Über die numerische auflösung von differentialgleichungen. Mathema- tische Annalen46(2), 167–178 (1895)

  25. [25]

    In: International Conference on Learning Representations (ICLR) (2026)

    Sarabia, R.P., Nyborg, J., Birk, M., Sjørup, J.L., Vesterholt, A.L., Assent, I.: Rainpro-8: An efficient deep learning model to estimate rainfall probabilities over 8 hours. In: International Conference on Learning Representations (ICLR) (2026)

    2026

  26. [26]

    Advances in Neural Information Processing Systems (NeurIPS)28(2015)

    Shi, X., Chen, Z., Wang, H., Yeung, D.Y., Wong, W.K., Woo, W.c.: Convolutional lstm network: A machine learning approach for precipitation nowcasting. Advances in Neural Information Processing Systems (NeurIPS)28(2015)

    2015

  27. [27]

    Advances in Neural Information Processing Systems (NeurIPS)30(2017)

    Shi, X., Gao, Z., Lausen, L., Wang, H., Yeung, D.Y., Wong, W.k., Woo, W.c.: Deep learning for precipitation nowcasting: A benchmark and a new model. Advances in Neural Information Processing Systems (NeurIPS)30(2017)

    2017

  28. [28]

    arXiv preprint arXiv:2003.12140 (2020)

    Sønderby, C.K., Espeholt, L., Heek, J., Dehghani, M., Oliver, A., Salimans, T., Agrawal, S., Hickey, J., Kalchbrenner, N.: Metnet: A neural weather model for precipitation forecasting. arXiv preprint arXiv:2003.12140 (2020)

    work page arXiv 2003

  29. [29]

    In: International Conference on Learning Representations (ICLR) (2026)

    Song, C., Chang, T.Y., Hong, Y.: Extreme weather nowcasting via local precipita- tion pattern prediction. In: International Conference on Learning Representations (ICLR) (2026)

    2026

  30. [30]

    arXiv preprint arXiv:2601.20342 (2026)

    Sun, H., Yang, Y., Han, W., Huang, W., Chen, H., Gao, Z., Li, Z., Huo, Z., Niu, Z.: Stormdit: A generative ai model bridges the 2-6 hour’gray zone’in precipitation nowcasting. arXiv preprint arXiv:2601.20342 (2026)

    work page arXiv 2026

  31. [31]

    IEEE Transactions on Multimedia (2025)

    Tan, C., Gao, Z., Li, S., Li, S.Z.: Simvpv2: Towards simple yet powerful spatiotem- poral predictive learning. IEEE Transactions on Multimedia (2025)

    2025

  32. [32]

    Transactions on Machine Learning Research (2026) 18 Y

    Tang, Y., Qi, L., Xie, F., Li, X., Ma, C., Yang, M.H.: Video prediction transformers without recurrence or convolution. Transactions on Machine Learning Research (2026) 18 Y. Seong et al

    2026

  33. [33]

    In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

    Tatsubori, M., Moriyama, T., Ishikawa, T., Fraccaro, P., Jones, A., Edwards, B., Kuehnert,J.,Remy,S.L.:Deeptemporalinterpolationofradar-basedprecipitation. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). pp. 1685–1689. IEEE (2022)

    2022

  34. [34]

    Advances in Neural Information Processing Systems (NeurIPS)33, 22009–22019 (2020)

    Veillette, M., Samsi, S., Mattioli, C.: Sevir: A storm event imagery dataset for deep learning applications in radar and satellite meteorology. Advances in Neural Information Processing Systems (NeurIPS)33, 22009–22019 (2020)

    2020

  35. [35]

    In: International Conference on Learning Rep- resentations (ICLR) (2024)

    Verma, Y., Heinonen, M., Garg, V.: Climode: Climate and weather forecasting with physics-informed neural odes. In: International Conference on Learning Rep- resentations (ICLR) (2024)

    2024

  36. [36]

    Expert Systems with Applications p

    Wang, H., Zeng, Q., Wang, H., He, J., Yu, T., Liu, G.: Temporal super-resolution reconstruction of weather radar echoes using a deep learning approach. Expert Systems with Applications p. 130189 (2025)

    2025

  37. [37]

    In: Proceed- ings of the International Conference on Machine Learning (ICML)

    Wang, Y., Gao, Z., Long, M., Wang, J., Yu, P.S.: Predrnn++: Towards a resolution of the deep-in-time dilemma in spatiotemporal predictive learning. In: Proceed- ings of the International Conference on Machine Learning (ICML). pp. 5123–5132 (2018)

    2018

  38. [38]

    In: International Conference on Learning Representations (ICLR) (2019)

    Wang, Y., Jiang, L., Yang, M.H., Li, L.J., Long, M., Fei-Fei, L.: Eidetic 3d lstm: A model for video prediction and beyond. In: International Conference on Learning Representations (ICLR) (2019)

    2019

  39. [39]

    In: Advances in Neural Information Processing Systems (NeurIPS)

    Wang, Y., Long, M., Wang, J., Gao, Z., Yu, P.S.: Predrnn: Recurrent neural net- works for predictive learning using spatiotemporal lstms. In: Advances in Neural Information Processing Systems (NeurIPS). vol. 30 (2017)

    2017

  40. [40]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Wang, Y., Zhang, J., Zhu, H., Long, M., Wang, J., Yu, P.S.: Memory in memory: A predictive neural network for learning higher-order non-stationarity from spa- tiotemporal dynamics. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 9154–9162 (2019)

    2019

  41. [41]

    Wen, Y., Schuur, T., Vergara, H., Kuster, C.: Effect of precipitation sampling error on flash flood monitoring and prediction: Anticipating operational rapid-update polarimetricweatherradars.JournalofHydrometeorology22(7),1913–1929(2021)

    1913

  42. [42]

    Advances in Neural Information Processing Systems (NeurIPS) 37, 100007–100041 (2024)

    Yan, C.W., Foo, S.Q., Trinh, V.H., Yeung, D.Y., Wong, K.H., Wong, W.K.: Fourier amplitude and correlation loss: Beyond using l2 loss for skillful precipi- tation nowcasting. Advances in Neural Information Processing Systems (NeurIPS) 37, 100007–100041 (2024)

    2024

  43. [43]

    Yoon, D., Seo, M., Kim, D., Choi, Y., Cho, D.: Probabilistic weather forecasting withdeterministicguidance-baseddiffusionmodel.In:ProceedingsoftheEuropean Conference on Computer Vision (ECCV). pp. 108–124. Springer (2024)

    2024

  44. [44]

    In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Yu, D., Li, X., Ye, Y., Zhang, B., Luo, C., Dai, K., Wang, R., Chen, X.: Diffcast: A unified framework via residual diffusion for precipitation nowcasting. In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 27758–27767 (2024)

    2024

  45. [45]

    In: International Conference on Learning Rep- resentations (ICLR) (2020)

    Yu, W., Lu, Y., Easterbrook, S., Fidler, S.: Efficient and information-preserving future frame prediction and beyond. In: International Conference on Learning Rep- resentations (ICLR) (2020)

    2020

  46. [46]

    Scientific Reports14(1), 19024 (2024)

    Zhao, Z., Wang, Z., Zhao, G., Zhao, J.: A new strong convective precipitation forecasting method based on attention mechanism and spatio-temporal reasoning. Scientific Reports14(1), 19024 (2024)

    2024